Charles Ryan: Hello, I'm joined today by Professor Alexander Wyatt who's an assistant professor at the Vancouver Prostate Center and the University of British Columbia. Alex and his team have published a really fascinating piece in nature that I think is going to help us understand prostate cancer a little bit better, if not a lot better, and it's certainly going to have clinical ramifications for how we think through this disease. Alex, thanks for joining us. I'm going to turn it over to you and have you give a brief PowerPoint presentation, and then we'll have a conversation.

Alexander Wyatt: Thank you very much Dr. Ryan and for Euro Today and the Prostate Cancer Foundation by hosting me. So, I am delighted to give a quick snapshot on our work and to have a discussion. So, just by way of introduction, circulating tumor DNA is short fragments of DNA in the blood. It mixes with cell-free DNA from non-cancer cells, mostly leukocytes, has a really short half-life, and it's associated with proliferative tumor burdens. So the more proliferative metastases you have in an individual with advanced disease, on average, the more ctDNA they have. And we've seen recently that ctDNA tests are actually suggestive of utility across a range of different clinical scenarios. Some of these are about detecting ctDNA. If you detect ctDNA, it can tell you about, for example, residual disease after surgery. Some of these uses are about characterizing ctDNA. So, if you have high levels of ctDNA, that can associate with poor prognosis. If you have a TP53 mutation, that could be poor prognosis. If you have a BRCA2 mutation, that could actually be predictable treatment benefit.

Alexander Wyatt: But I think there are several remaining outstanding questions about just fundamental ctDNA biology that were not answered yet, and these were the questions we wanted to address. So firstly, does ctDNA actually harbor multiple populations. This is kind of an often hypothesized thing, but never really been proven. And if these populations are present, are they changing in prevalence during treatment? Are they evolving? And finally, what extra information can we get about these different ctDNA populations, not just genome changes. So to do this, we needed to generate deep whole genome sequencing of serial clinical samples, and no prior data set existed. So, we teamed up with professor Felix Feng at the University of California, who has had the privilege of working on the PCA stand-up to cancer dream team cohort, which includes many patients that generously donated their metastatic tissue and provided a blood sample at the same time. And in parallel, we collaborated with Kim Chi here in Vancouver, and collectively, we generated this data set of deep whole genome sequencing of ctDNA, including patients that provided serial samples over time, as well as those with matched metastatic tissue.

Alexander Wyatt: The first major finding that we reported was that ctDNA is indeed much more diverse than a single metastatic lesion. There are many different ways to measure this. This particular image is telling us that there's diversity in copy number changes. And what we were able to do is prove that each individual metastasis actually only contributes a minor share of total ctDNA. So indeed, ctDNA is comprised of different populations shedding into the bloodstream.

Alexander Wyatt: Now, this is all very well and good, but can you actually, when you know there's multiple populations present, actually deconvolute these to understand what each one is doing? And so we wrote a series of computer software of algorithms to be able to do this, to pull apart the different populations that are present. And what we found is that there is a range of different scenarios, some of which that are very complex, but on average, we see that there are these different populations in present prior to treatment, and some of these populations expand after treatment, and these can be subtly different. And for example, what we saw is that patients with BRCA2 mutations or mismatch repair defects. We see continued accrual of defects associated with those underlying instability, so indicating that say BRCA2 defect is still driving accrual of new mutations. And ultimately, this is telling us really the highest level of evidence so far that evolution of populations in these patients is still going on, even in the advanced CRPC setting.

Alexander Wyatt: So, this is great, but why would this matter from a clinical perspective? What key resistance mechanisms are coming through? And we found that in patients that are receiving AR targeted therapy when we look across the whole genome at copy number changes in mutation and structural rearrangements, there's really only one area that's consistently changing, and that was the androgen receptor. And so this is telling us that the populations that are being selected, even in the CRPC setting, are still those with increasingly aggressive androgen receptor alterations. And to give you an example of what this looks like, one patient where we see emergence of a new population, this green clone that's expanding, this actually carried 50 copies of the AR, compared to before treatment when we saw sort of 20 to 30 copies, so really showing how you can select out increasingly aggressive AR alterations.

Alexander Wyatt: So finally, we wanted to look at phenotype, phenotype information. And we know that, obviously, you can do methylation profiling of ctDNA to kind of get a profiles, but we wanted to see whether you could get some inference about phenotype from the DNA sequencing data itself, from the whole genome sequencing data we've already got. And this is plausible because ctDNA and cell-free DNA in general comes pre-fragmented by the apoptotic machinery of the cells. It's not like when we get a piece of tissue and we have to chop that DNA up ourselves in order to sequence it. The cancer's already done it for us. So, we can actually then infer where nuclear zones in the cell were, because the apoptotic machinery does preferentially cut in between nuclear zones. So in theory, you can understand gene transcription and transcription factor binding from ctDNA itself. And this is in fact what we saw.

Alexander Wyatt: So, to highlight two specific things, firstly, we could actually see that highly expressed genes that we knew were expressed in a patient's tumor tissue, we could see features consistent with high expression in ctDNA, and genes that were lowly expressed did not show these features. And again, we wanted to come back and say, well, why could this matter? What information could you actually pull out that could be of clinical relevance? So, we turned to the androgen receptor and looked across all these different AR binding sites in the genome, so 3000 different regions where androgen receptor binds. And what we looked for is evidence, essentially, that AR was bound to DNA at those regions. The vast majority of samples, we see clear evidence for that. That's represented by these sort of red or yellow regions in the middle, but you can see a big range here. And actually, those samples, which showed the highest level of AR occupancy at binding sites, they tended to have AR mutations or high AR applications. And the patients with low levels of AR occupancy, that included those with neuroendocrine prostate cancer.

Alexander Wyatt: So, this is potentially telling us that we may, from DNA sequencing, be able to understand the level of AR activity in a cell. And the final piece of information I'll show is a sort of anecdote that gives a proof of principle for this, where a patient with bone predominant castration-resistant prostate cancer, at baseline, after two lines of treatment, we actually see expansion of a new population in the ctDNA that's carrying an RB1 mutation, and we see the complete aggregation of the AR binding site occupancy in the ctDNA. And this patient subsequently progressed with liver metastases, and then biopsy confirmed neuroendocrine disease. So together, this is showing us, from the same piece of sequencing data, we can infer the genomic kind of enablers and actually see the adaptive changes happening live. So in future, we can apply this technology to give deep insight into temporal genotype and phenotype in patients, and I think that's where the blood test technology is really powerful to look when we can solve the patients over time.

Alexander Wyatt: And we think this is relevant for latest emerging therapies, Lutetium PSMA and others, where we don't yet understand the resistance mechanisms. And ultimately, I think this could help develop a multimodal liquid biopsy where, from the same test, we can get lots more than just genomic information. So, I will stop there and unshare. Share my screen.

Charles Ryan: Alex, thank you. Impressive effort, massive team effort, as you just showed on your last slide with many groups from around the multiple countries working on this. As I'm looking at this, I'm seeing a few key takeaways that has some clinical relevance. One is, unlike doing a biopsy of a metastasis, when you examine the circulating tumor DNA, you're seeing the full mixture of all of the potential genomic alterations that can be there. And that's just something that you there's really no other way to capture all of that in real time. Number two, you can capture that seriously, so you're able to look at changes over time. And then finally, the two other points that I think are interesting are you come back to the androgen receptor. It's still the most important alteration we see. Despite all we know about cancer biology and all these other genes and all these other interactions, the energy receptor, I think you said something 70 some mutations or something in some samples.

Charles Ryan: And then you really drive that home by looking at the AR occupancy of the nucleosome occupancy. So, not only is the androgen receptor there, it's mutated, but it's where it is going to be when it's exerting its effect. And so this enables a whole series of changes in how we might conceive of the disease in the clinic and how we might test it. How do you see this getting to the clinic in the coming year or so? What should clinicians, what I do in my clinic when I'm thinking about ctDNA with the existing panels that exist? And how might that evolve?

Alexander Wyatt: So, I think the existing panels that are commercially available at the moment, they don't really provide any of the tipping information. They can obviously help you understand how androgen receptor is changing over time. So, if you see kind of a change in AR genotype between one sample and the next, I think that might actually be telling you that that tumor is still somewhat reliant on the AR. I do think we're going to see a rapid translation of the fragmentary phenotypic information to existing panels. And that's one thing we are trying to explore. Whole genome sequencing is not feasible to do on every patient, right? But in theory, we should be able to find the key pieces of information and put them in a panel. And that's something we are kind of actively working on at the moment. There are these new generations of AR inhibitors coming through at the moment in preclinical development, and we need biomarkers for identifying patients that may still be particularly addicted to the AR signaling pathway. And to me, this seems like one potential option for that, but obviously, you guys are the experts.

Charles Ryan: Well, it's a great question because one could say... You could argue this both ways. You could say, "Well, we see that the AR continues to be public enemy number one, so let's continue to think about novel ways we can target the androgen receptor." Alternatively, you can say, "Well, look at what happened. The more and more we suppress the androgen receptor, look at what happens. RB1 is lost, and it becomes neuroendocrine and it kills the patient." And so one could basically surrender the AR and say, "We're doing as well as we can. We need to focus on other mechanisms." And I think that's a really interesting point you made about RB1 and that really nice diagram you have where the RB1 loss proportion just went up, and then this patient developed a much more aggressive clinical phenotype. And so I think we need need to attack both.

Charles Ryan: Yes, I would love to see better AR targeted therapies, but maybe that's not the answer to the whole big picture. And I think that that's one of the keys to this. Another thing that I think is really important that you show is, at least in terms of that diagram, the RB1 loss is something of almost a continuous variable. This is more of a question now, is there a moment in the disease where RB1 is lost or some other alteration occurs, and the entire tumor burden switches to that phenotype? Probably not. Right? And I think what you're showing is that RB one was lost in a proportion. And it grows out to become the predominant tumor burden in the patient. Is that correct assessment?

Alexander Wyatt: Yeah, I think there are some alterations, obviously, that arise really early in cancer development, typically DNA repair defects in one of those. And so they're kind of present in every daughter cell, and there's not much you can do to change that. Things like RB1 alterations, we know arise relatively late and can still arise in late-stage disease. So, it's only then when we apply the selective pressures of drugs that we can actually encourage, or at least we provide the environment where those have an advantage over the other clients. And so I think that's what we're seeing. The vast majority of tumors don't develop clones that have RB1 deficiency, and so we don't see those grow out. But in those that do, that does then obviously lead to a particularly aggressive scenario. But I think it tells us that even once you have a patient, once you have an individual that's developed castration-resistant disease, that genotype and phenotype is not locked in, and we can continue to see evolution going on, especially now that we are actually able to use multiple drugs, now that we have all these options available.

Alexander Wyatt: And so to me, it suggests probably we still need to be monitoring the disease in the late stage, and I think that's where blood gives you an opportunity to be inclusive and profile patients repeatedly over time.

Charles Ryan: Well, it's fascinating as a clinician and as somebody who thinks about this science, that with all of our new therapies, we help patients, but we create new diseases at the same time. And those new diseases are even more virulent than the one we were treating before. People didn't have RB... Very few people had RB one deficient prostate cancer in the 1990s, but now that we have other new therapies, it's becoming more and more prevalent. So, that's a challenge, but it's also, of course, an opportunity for us at the Prostate Cancer Foundation and you laboratories and elsewhere around the world to demonstrate novel approaches to this problem, which applies obviously to prostate cancer, but to other cancers as well. And I think that's a unique opportunity that these data may present to us.

Charles Ryan: One question as I was looking at this, and I was thinking about the RB1 loss or the emergence, is it possible, do you think, if we were to look at this serially, in a greater degree of granularity, that we could identify factors that precede or almost herald the onset of RB1 loss so that we could look at these cases and we could say, "Well, these are the factors that were associated with the eventual emergence of this"? Because as you point out, you could put 50 patients on enzalutamide and only a handful are going to have RB1 loss. And we think we know it's because of high-grade tumors, et cetera, et cetera. But what are the events that precede that, or the events that predict that? Do we have any sense from your data?

Alexander Wyatt: I think that's not something we can directly answer with our data so far. I do agree there probably are some sort of environments where some genotypes, let's say, where something like an RB1 alteration is more likely to occur. We probably need a lot more samples, a lot more patients, a lot more disease to be studied in order to find those correlations. But you're probably right that there are scenarios where there's almost a limitation on what types of things can accrue. I mean, one concept would be a tumor that has developed acquired 50 copies of the receptor is very, very well down that line. And we may not expect those types of tumors to then commit to coming all the way back and exhibiting lineage plasticity. So, I think there are perhaps some clues. Where probably our data can help is that we can spot these clones when perhaps they're in the minority. So prior to treatment, those populations may actually be present in a not representative of bulk disease. They may not be spotted in a single metastasis, but they're there in circulation and representative of some lesion somewhere in the body.

Alexander Wyatt: And then we may know, okay, this is the type of situation where when we really raise the fitness threshold on AR, we may actually create an environment where these other clones can outgrow, but perhaps we don't have any alternatives to target them right now anyway.

Charles Ryan: Yeah, it's a fascinating paper and really interesting insights into the disease. Just as an aside, do you see, in the literature, this technology showing similar applications to other cancers? Do we see this type of dynamics with regards to the emergence of clones and ctDNA and other cancers?

Alexander Wyatt: So, this was really the first of its kind across all cancers, in the sense that there'd never been a deep whole genome study in any cancer type for ctDNA, but we really do recognize that there are many other cancers that have high levels of ctDNA, metastatic cancers. Even breast cancer has high levels of ctDNA in the advanced setting. So, we actually made all of our methodology open source. The code is publicly available. The data's publicly available. So, we think that this has translational relevance for all types of other cancers, and we are really trying to encourage other researchers to leverage our methods because that was a big aspect. We had to develop these new tools to be able to study the data. So I think whilst this research started in prostate, it is going to influence all other types of cancers that share ctDNA.

Charles Ryan: And I would agree. And that's one of the reasons why we're so proud that the PCF could have our fingerprints on this a little bit, and proud to have you amongst our investigators. So, to summarize, as a clinician, I think that these data are fascinating from a couple of different perspectives. And number one, they demonstrate that circulating tumor DNA is likely more informative than doing a tumor biopsy, and that's something that the field is going to have to sort out over time. It's much easier to do a blood draw and to identify DNA than it is to biopsy a bone, certainly, but I think that hopefully, the integration of circulating tumor DNA into clinical care will happen over time. A key limitation of this study is that they did, as Dr. Wyatt, pointed out a whole genome sequence, and that's not really feasible in an everyday clinic type of environment.

Charles Ryan: So, what will have to happen is, we'll have to figure out, as a field, what are the most distilled important parts to include on future panels that could be done as a part of this, to look at prostate cancer. And then finally, I think that the key is it would be really great if these data and similar data can help show the world that we have emerging phenotypes, that the disease evolves as we treat it. And we think of the disease as sensitive or resistant to a particular drug. And that's very easy from a clinical perspective, but really sensitive to enzalutamide or abiraterone versus resistant with RB loss, versus resistant with an intact RB versus resistant with TP53 mutations or resistant with 70 AR mutations, those are essentially different diseases. And so I think that's one of the key clinical features that we have to respond to over time. And I want to thank you for all this effort and this contribution, not only to prostate cancer but to science as well. So, Alex Wyatt, thank you so much for joining us.

Alexander Wyatt: Thank you, Dr. Ryan.